This study employed a cross-sectional survey design. Between September 4 and September 30, 2024, a convenience sampling method was implemented to distribute an online questionnaire survey targeting potential participants for the national PEE in the cities of Nanchang, Fuzhou, Wuhan, and Hangzhou. Participants were required to meet specific inclusion criteria: they needed to express an intention to participate in the 2025 national PEE, provide informed consent, and engage in the survey voluntarily. The exclusion criteria encompassed individuals who did not intend to participate in the 2025 national PEE and those with anxiety disorders or other emotional disturbances. The recruitment process rigorously adhered to the principles of voluntary participation and informed consent. All participants confirmed their voluntary engagement through an electronic informed consent procedure prior to accessing the survey.

Data were collected anonymously utilizing the Questionnaire Star platform¹, which is a widely employed tool in China for data collection, survey administration, and various research purposes. This platform offers a user-friendly interface and comprehensive functionalities for designing, distributing, and analyzing questionnaires. A QR code was generated to facilitate easy dissemination, linking directly to the content of the questionnaire. Our research team strategically employed online social platforms and relevant social resources to distribute the questionnaire. Participants completed the questionnaire after providing informed consent, and measures were implemented to ensure that each IP address could submit the questionnaire only once, thereby preventing potential duplications and safeguarding data integrity. The questionnaire included standardized survey instructions, mandating that all questions be answered to uphold respondents’ anonymity and confidentiality.

Based on the sample size requirements for exploratory factor analysis (EFA) as outlined in reference (Wu, 2010), a minimum target of 500 participants was established. This target was calculated as 5–10 times the number of questionnaire items (k = 46), resulting in a range of 230–460 participants, with additional provisions made for potential invalid responses. Ultimately, a total of 515 questionnaires were collected. To ensure data quality, dual-stage quality control procedures were implemented, which included the exclusion of responses with completion times less than the mean minus three standard deviations (i.e., less than 93 s), and the elimination of patterned responses characterized by 80% or more items receiving identical answers. These procedures resulted in 466 valid samples, yielding a valid response rate of 90.49%.

Among the 466 valid participants, 46 individuals (9.87%) who reported no engagement with knowledge-based videos were classified as the “non-engagement” group, while the remaining 420 participants (90.13%) with varying levels of engagement were categorized as the “engagement” group. A chi-square test was performed to assess the homogeneity of demographic variables between the engagement and non-engagement groups. The results indicated that none of the variables reached statistical significance (p > 0.05), suggesting the absence of systematic distributional differences in demographic characteristics between the groups. This finding supports the conclusion of balanced baseline characteristics between the two groups. Table 1 presents the basic demographic information of participants in each group.

Data analysis was conducted utilizing SPSS version 23.0. A paired chi-square test was employed to assess the homogeneity of participants between the engagement and non-engagement groups. The Kolmogorov–Smirnov (K-S) test was utilized to perform normality testing on the data, while Harman’s single-factor method was applied to evaluate potential common method bias in the collected data. To examine differences in learning anxiety and self-efficacy among PEE candidates with varying levels of knowledge-based videos engagement, an analysis of variance (ANOVA) was performed, followed by post hoc multiple comparisons using the Least Significant Difference (LSD) method. A partial correlation analysis was conducted to investigate the relationships among knowledge-based videos engagement, self-efficacy, and learning anxiety. Furthermore, a mediation analysis was conducted employing Hayes’ PROCESS macro, which utilized the Bootstrap resampling method with 5,000 replications. Significance was evaluated through the application of 95% confidence intervals.

In this study, the Kolmogorov–Smirnov (K-S) test was utilized to evaluate the normality of the questionnaire data. The findings revealed that the absolute values of kurtosis for professional identity, self-control, and unethical professional behavior were 0.816, 0.587, and 1.105, respectively, while the absolute values of skewness were 0.163, 0.072, and 0.495, respectively. Drezner and Zerom (2010) contend that the criteria for normality tests are typically stringent, and absolute normality is seldom attained. According to established guidelines, if the absolute value of kurtosis is less than 10 and the absolute value of skewness is less than 3, the data can generally be regarded as conforming to a normal distribution. Consequently, the data in this study can be reasonably accepted as adhering to a normal distribution.

Common Method Bias is a phenomenon encountered in research when identical or similar methods are employed for data collection. This can lead to an artificial and misleading correlation among data points, which does not accurately reflect a genuine relationship between variables. Instead, it represents a bias arising from the limitations or consistency of the research methods (Podsakoff et al., 2003). In this study, we utilized Harman’s single-factor method to evaluate the presence of common method bias in the questionnaire data. The results indicated a χ²/df ratio of 3.82, exceeding the threshold of 3. Additionally, the Goodness of Fit Index (GFI), Comparative Fit Index (CFI), Normed Fit Index (NFI), and Non-Normed Fit Index (NNFI) values were 0.86, 0.84, 0.80, and 0.83, respectively, all falling below the desired level of 0.90. The Root Mean Square Error of Approximation (RMSEA) value was 0.18, surpassing the critical standard of 0.10, while the Root Mean Square Residual (RMR) was 0.09, deviating from the reference value of 0.05. Factor analysis revealed that the variance contribution rate of the first factor was 37.86%, which is below the expected critical value of 40% (Podsakoff et al., 2003). This suggests that the data from the questionnaire in this study could not be consolidated into a single factor, indicating good discriminability in the measurements. Therefore, there was no significant common method bias in the research data, and it is unlikely to have adversely affected the results of this study.

This study faced methodological challenges in evaluating the levels of qualitative variables. The lack of normative references in the design of the questionnaire precluded the use of traditional standardized comparisons for the interpretation of results. To mitigate this issue, the study employed the theoretical median reference approach, a method commonly utilized in the social sciences (Podsakoff et al., 2012), thereby establishing a conceptual evaluation framework. This approach leverages the theoretical median of the questionnaire items as a conceptual threshold to facilitate a comparative analysis of the measured means of knowledge-based videos engagement, learning anxiety, and self-efficacy. This methodologically adaptive strategy provided a theoretically robust interpretive framework in the absence of empirical anchoring data.

The findings indicated that the mean score for knowledge-based videos engagement among PEE candidates was 3.64 ± 1.25, which exceeds the theoretical median of 3 (as determined by the mean Z-test, z = 9.35, p < 0.001). Additionally, 83.69% of participants reported experiencing “moderate, severe, or extremely severe knowledge-based videos engagement during their exam preparation. These results suggest that PEE candidates exhibit a relatively high level of knowledge-based videos engagement. The average score for learning anxiety among PEE candidates was 3.23 ± 0.87, also surpassing the theoretical median of 3 (as indicated by the mean Z-test, z = 5.17, p < 0.001), which implies elevated levels of learning anxiety. Finally, the average self-efficacy score was 2.86 ± 1.11, which is slightly below the theoretical median of 3 (as shown by the mean Z-test, z = −4.38, p < 0.001), indicating moderate to somewhat low levels of self-efficacy among PEE candidates.

In the realm of multivariate data analysis, the correlation between two variables may be confounded or distorted by the influence of additional variables. Partial correlation analysis, as described by Podsakoff et al. (2003), addresses these confounding factors, thereby providing a more precise evaluation of the direct relationship between the variables under investigation.

As indicated in the preceding general information questionnaire, demographic variables can significantly influence key variables. Therefore, this study conducts a partial correlation analysis involving 420 PEE candidates from the engagement group. Seven demographic variables—specifically, gender, age, residence, type of school, major, interdisciplinary (Yes/No), and target institutions—are incorporated as control variables within the model. This approach aims to elucidate the intrinsic relationships among participation in knowledge-type videos, self-efficacy, and learning anxiety.

The findings indicated a positive correlation between knowledge-based videos engagement and learning anxiety (r = 0.19, p < 0.01), as well as a negative correlation between knowledge-based videos engagement and self-efficacy (r = −0.22, p < 0.01). Additionally, self-efficacy exhibited a negative correlation with learning anxiety (r = −0.48, p < 0.01). A summary of these results is presented in Table 4.

The study conducted an examination of the mechanisms underlying the relationship between knowledge-based videos engagement and learning anxiety among PEE candidates. It identified knowledge-based videos engagement as the independent variable, learning anxiety as the dependent variable, and self-efficacy as the mediating variable. After controlling for demographic factors as covariates, the Process program was utilized to analyze the mediating role of self-efficacy. This program, developed by Andrew F. Hayes, is a plug-in specifically designed for mediation and moderation analysis, enhancing the capabilities of traditional statistical software such as SPSS and SAS (Wen and Ye, 2014).

The results are presented in Table 5 and Figure 2. Model 1 demonstrated that knowledge-based videos engagement had a significant positive predictive effect on learning anxiety (β = 0.19, p < 0.001), with a 95% confidence interval of [0.11, 0.27] that did not include zero. This finding indicates that for every one standard deviation increase in engagement, learning anxiety levels increased by an average of 0.19 standard deviations, with the model accounting for 14% of the variance (R² = 0.14). Model 2 further revealed that knowledge-based videos engagement significantly and negatively predicted self-efficacy (β = −0.21, p < 0.001; 95% CI [−0.31, −0.11]), suggesting that higher engagement was associated with lower self-efficacy. The explanatory power of the model increased to 26%. When both knowledge-based videos engagement and self-efficacy were included in Model 3, self-efficacy exhibited a strong and significant negative predictive effect on learning anxiety (β = −0.50, p < 0.001; 95% CI [−0.56, −0.44]), while the direct effect of engagement became non-significant (β = 0.08, p = 0.051; 95% CI [−0.00, 0.16]). The overall explanatory power of the model rose to 33%.

The stability of the full mediation effect was further validated through the Bootstrap method. As presented in Table 6, the direct effect of knowledge-based videos engagement on learning anxiety was measured at 0.08, with a 95% confidence interval of [0.00, 0.14], which includes zero, indicating a lack of statistical significance. In contrast, the mediating effect of self-efficacy between engagement and learning anxiety was found to be 0.11, with a 95% confidence interval of [0.06, 0.16], which excludes zero, thereby confirming its statistical significance.

The findings indicate that, due to the non-significant direct effect and the significant mediating effect, self-efficacy serves as a complete mediator in the relationship between knowledge-based videos engagement and learning anxiety among PEE candidates, accounting for 100% of the total effect. This outcome implies that knowledge-based videos engagement indirectly intensifies learning anxiety by significantly diminishing self-efficacy, while the direct effect is entirely mediated through this pathway.

This study illustrates that PEE candidates exhibit a relatively high level of knowledge-based videos engagement. This finding is consistent with the research conducted by Chen and Lin (2024) and Wang et al. (2023). The observed engagement may be attributed to the candidates’ substantial demand for learning materials, the time constraints they face during their preparation, and their awareness of the online learning resources available to them. Given the rigorous academic and research requirements associated with the PEE, candidates necessitate extensive learning resources to ensure adequate preparation. Knowledge-based videos, as a form of multimedia learning tool, provide intuitive and engaging content that enhances comprehension and mastery of the subject matter. Moreover, due to the typically limited preparation time, candidates must adopt efficient study and review strategies, rendering knowledge-based videos essential for fulfilling their educational needs, as these resources effectively convey information. As we progress further into the digital media era, the proliferation of knowledge-based videos on online platforms is increasingly becoming integrated into individuals’ daily lives and learning environments, thereby enriching the online learning experience for PEE candidates who predominantly rely on self-directed study. Consequently, PEE candidates exhibit significant knowledge-based videos engagement throughout their preparation process.

This study further elucidates that PEE candidates experience considerable learning anxiety, a finding that is consistent with the research conducted by Gao (2019) and Wang et al. (2023). These findings suggest that the substantial pressure to succeed in the PEE, stemming from candidates competing against a large cohort for limited opportunities, leads to considerable psychological stress and anxiety. Moreover, the broad scope of PEE preparation exacerbates this anxiety, compounded by a lack of clear examination requirements and insufficient information regarding their learning progress and the performance levels of their peers (Liu, 2022; Wang et al., 2023). Whether they are students or graduates, PEE candidates are also required to balance their academic obligations with their primary employment during the preparation period. Ineffective time management can lead to an increased mental burden for PEE candidates, thereby heightening their anxiety levels.

According to the study, moderate to high levels of knowledge-based videos engagement were found to increase learning anxiety among PEE candidates, which is consistent with the research findings of Li and Guo (2023) and Li and Zhang (2024). This finding supports both the Information Overload Theory (Eppler and Mengis, 2010) and the Cognitive Load Theory (Sweller, 2010). As previously noted, knowledge-based videos engagement is crucial for PEE candidates to fulfill their learning requirements (Chen and Lin, 2024). When engagement with these videos is low, candidates may lack adequate learning resources to prepare effectively for the examination, potentially resulting in heightened anxiety. Conversely, reduced engagement allows candidates to absorb information more readily, fostering a sense of control and greater confidence in their preparation, which may lead to no significant increase in anxiety levels. As knowledge-based videos engagement increases, so too does learning anxiety among PEE candidates. From the perspective of Information Overload Theory (Eppler and Mengis, 2010), moderate to high levels of engagement may signify a potential overload, wherein candidates are inundated with a considerable amount of complex information within a brief period, complicating their ability to effectively filter and assimilate it. This situation can result in information overload, which subsequently induces stress and anxiety. Furthermore, it may lead to choice anxiety, wherein candidates experience apprehension when selecting specific videos from a diverse array of learning resources, fearing they may overlook critical information or select inappropriate materials. This uncertainty further exacerbates learning anxiety. From the perspective of Cognitive Load Theory (Sweller, 2010), when PEE candidates engage with numerous knowledge-based videos—especially those that present complex content or material that surpasses their current understanding—they are required to allocate additional cognitive resources to effectively process the information. This increased cognitive demand elevates the intrinsic cognitive load. If the instructional strategies employed in the videos are inadequate, such as lacking sufficient explanations, examples, or practical exercises, learners may struggle to integrate new knowledge into their existing cognitive frameworks, thereby intensifying the cognitive load. Prolonged or excessively high cognitive load can lead to increased stress among learners, consequently heightening their learning anxiety. They may perceive the learning tasks as more complex and challenging. If their personal cognitive and coping resources are insufficient to manage this load, anxiety levels will rise significantly. Conversely, with limited knowledge-based videos engagement, candidates can more easily assimilate the information, resulting in a weak correlation between video engagement and learning anxiety.

This study, through a comprehensive analysis of data, demonstrates that there is no statistically significant direct association between knowledge-based videos engagement of PEE candidates and their levels of learning anxiety. Furthermore, self-efficacy is shown to fully mediate this relationship. These findings not only offer empirical support for research concerning emotional mechanisms within digital learning environments but also align closely with the foundational principles of the Social Cognitive Theory (Bandura, 1985), elucidating the psychological pathways through which learning behaviors impact emotional states.

From the perspective of the Social Cognitive Theory, the findings robustly support the central proposition that inputs require cognitive mediation to exert their effects. Knowledge-based videos, as quintessential stimuli within digital learning environments, possess dual characteristics: they present intellectual challenges through complex concepts while simultaneously posing the risk of information overload. However, such environmental stimuli do not directly dictate emotional responses; rather, they must be filtered and transformed through individual cognitive systems. Specifically, when prospective educators engage with knowledge-based videos, their self-efficacy—serving as a critical cognitive mediator—governs the appraisal and interpretation of these environmental stimuli. Candidates who successfully comprehend and internalize the content of the videos enhance their self-efficacy beliefs through mastery experiences, a positive cognitive restructuring process that regulates emotional states and mitigates learning anxiety. Conversely, when individuals perceive the video content as exceeding their capabilities, negative self-efficacy evaluations exacerbate anxiety through physiological arousal and behavioral avoidance. This dynamic interaction illustrates that external factors influence emotional outcomes solely through their subjective interpretation as mediated by self-efficacy (Mclean et al., 2014).

The identified full mediation effect carries significant theoretical and practical implications. From a theoretical standpoint, the role of self-efficacy as a central nexus linking external learning behaviors and internal emotional states offers a novel perspective for understanding learning anxiety in the digital age. This perspective suggests that anxiety fundamentally reflects learners’ cognitive appraisals of their adaptability to digital environments, rather than merely the intensity of media exposure (Lau and Lee, 2020). Practically, these findings inform educational interventions that employ dual-path strategies: cognitive restructuring, such as the implementation of scaffolded learning objectives to enhance self-efficacy, and environmental optimization, for instance, the utilization of adaptive algorithms to regulate video information density. These approaches can assist learners in establishing constructive self-assessment mechanisms during human-technology interactions, effectively disrupting the detrimental cycle of information overload, eroded self-efficacy, and heightened anxiety. Ultimately, this can promote emotional well-being and enhance academic performance within digital learning ecosystems.

This study empirically investigates the relationship between knowledge-based videos engagement of PEE candidates and their levels of learning anxiety, while also elucidating the mediating role of self-efficacy in this relationship. The findings not only enhance the understanding of Social Cognitive Theory but also offer evidence-based recommendations for PEE candidates to effectively leverage knowledge-based videos in their learning processes. The principal conclusions are as follows:

Sample size and geographic limitations. The sample size of this study is relatively small compared to the overall population of PEE candidates, and the geographic concentration of participants in Eastern China may limit the generalizability of the findings. Future research should employ multi-regional stratified sampling to enhance external validity.

The increased prevalence of knowledge-based videos engagement, in conjunction with the widespread incidence of learning anxiety among PEE candidates, underscores the pressing necessity for targeted interventions. Our research indicates that knowledge-based videos engagement is a significant predictor of learning anxiety, particularly at moderate to high levels of engagement, with self-efficacy acting as a crucial mediating variable.

In light of the research findings presented, we propose that the core components of Social Cognitive Theory (Bandura, 1985) be integrated with contemporary advancements in Cognitive Load Theory (Sweller, 2010) for this particular population. The development of a three-dimensional intervention model encompassing “environment-cognition-behavior,” this approach aims to offer both theoretical guidance and practical strategies for mitigating educational anxiety in the digital era. The following specific recommendations are made.

First and foremost, it is crucial to balance engagement with the use of knowledge-based videos. Promoting the judicious application of knowledge-based videos is essential. While it is acknowledged that such videos offer significant advantages for learning and preparation, it is equally important to remain vigilant regarding the potential for excessive engagement to exacerbate anxiety. To mitigate this risk, it is advisable to establish a daily viewing limit (Lu, 2022) and prioritize high-quality content from reputable sources (Li et al., 2023; Li and Zhang, 2024). Furthermore, students should be encouraged to develop structured and attainable study plans, as well as to employ a variety of learning methods to accommodate diverse learning styles (Pekrun and Stephens, 2009). It is imperative to avoid an overreliance on knowledge-based videos engagement. Additionally, providing media literacy training is crucial, as it equips students with the skills necessary to critically evaluate the content they consume. Understanding algorithmic mechanisms and recognizing media biases can empower students to prevent feelings of overwhelm and mitigate any adverse effects stemming from the content to which they are exposed (Liu, 2024).

Secondly, it is imperative to implement interventions designed to enhance self-efficacy among students. Initiatives aimed at bolstering students’ self-efficacy should be developed, which can be accomplished through the organization of seminars, lectures, and personalized tutoring sessions, all with the objective of fostering students’ confidence in their academic endeavors (Sasson and Tifferet, 2025). Furthermore, the incorporation of techniques such as goal-setting, positive self-talk, and visualization may also contribute to the improvement of self-efficacy (Xie et al., 2021).

A comprehensive strategy for reconstructing the learning environment to mitigate learning anxiety should be implemented. The integration of stress management techniques is essential for alleviating anxiety. It is advisable to incorporate practices such as mindfulness meditation, yoga, and deep-breathing exercises into daily routines, as these methods have been shown to have a positive impact on anxiety reduction (Qiu, 2022). Additionally, promoting regular physical activity can significantly enhance emotional regulation (An and Peng, 2024). The establishment of peer support networks is also crucial; such networks enable students to share experiences and strategies, fostering mutual encouragement. Peer-to-peer learning can notably diminish anxiety and cultivate a sense of community among students (Wang and Wang, 2024). Furthermore, it is imperative to ensure that students have access to professional counseling services and to encourage them to seek assistance when feeling overwhelmed, as counseling can provide effective resources for managing anxiety and stress.

By employing these strategies, educational institutions can enable PEE candidates to proficiently manage their engagements with knowledge-based videos, improve their self-efficacy, and alleviate learning anxiety, thus promoting a healthier and more productive learning environment.